A method for performing region growing in a data volume that accounts for noise
in the data on a probabilistic basis. The data volume is divided into discrete
cells, and a probability distribution for each cell's datum value is calculated.
Each probability distribution is randomly sampled to generate one probabilistic,
noise-free realization for the data volume, and region growing is performed using
selected criteria. The process is repeated for different realizations until sufficient
statistics are accumulated to estimate the probable size and connectivity of objects discovered.
